This article is concerned with Model Predictive Control (MPC) algorithms that use Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks for prediction. For two benchmark processes, it is shown that the typical approach to MPC that hinges on successively linearized LSTM or GRU models do not give precise predictions and satisfactory control quality. The presented MPC control schemes utilize online advanced trajectory linearization, which yields simple quadratic optimization programs. It is shown that the discussed approaches give excellent prediction accuracy and control quality, very similar to that possible in MPC with full nonlinear prediction and nonlinear optimization done online. It is also demonstrated that the described MPC algorithms are a few times faster than the MPC method with nonlinear optimization. Moreover, the performance of MPC based on LSTM and GRU networks is compared, and simpler GRU networks are recommended.

The classical approach to thermal comfort control utilizes fully automatic controllers for temperature, humidity, and other variables. Although commonly used, such systems may need much energy. If speed is not the priority, natural airflow may be considered for the control process. However, this would require a model including airflow changes caused by the opening of windows and doors, which has not been found in the literature yet. This article presents a novel multi-layer control system for thermal comfort control that uses an original graph-based representation of opening and closing doors and windows. It allowed us to model airflow dynamics. The system has two essential advantages. Firstly, the natural airflow between adjacent zones may help the automatic controllers achieve their goals and save energy. Secondly, including the windows and doors into the model lets the human play an active role in an eco-aware control process, which corresponds with the “human-in-the-loop” trends. The concept has been illustrated with an example house with four types of thermal comfort zones. The optimization approach based on finding the optimal subgraph of opened windows and doors between chosen zones led up to 5% energy savings of the electric actuators, compared with the classical fully automated structure.